Traditional preference learning models cannot deal with uncertain data, while the fuzzy support vector machine can overcome the influence of uncertain data on learning effects.This paper studies the preference learning models based on the fuzzy support vector machine.Specifically, a linear-ranking method based on the fuzzy support vector machine and preference learning is proposed to address the complete ranking problems.Meanwhile, an ordinal regression method based on the fuzzy support vector machine and preference learning is proposed to address the classification-ranking problems.Considering decision-makers' sensitivity to classification errors, a loss-sensitive fuzzy membership function is proposed.Validity and rationality of the proposed methods are confirmed through a case study focused on predicting the severity of Parkinson's disease.Comparative analysis shows that the two methods have lower classification error or higher classification accuracy in predicting the severity of Parkinson's disease.The preference learning models with fuzzy support vector machine can overcome the drawback of traditional fuzzy support vector machine methods and reflect the uncertainty in decision processes, extending the range of applications for machine learning models.
为了处理分类问题中的不确定信息,学者们将模糊理论引入支持向量机并提出了模糊支持向量机(Fuzzy Support Vector Machine,FSVM)[18]。学者们构建了不同的模糊隶属函数来处理不确定信息。杜喆等[19]提出了一种基于类超平面的模糊隶属函数构造方法,该方法将聚类中心作为类超平面上的固定点,并将相邻两类中心的连线作为法向量。Heo和Gader[20]利用主成分分析法来实现噪声数据的平滑处理从而不再要求数据分布特征的先验知识。为了克服不同模糊隶属函数构造方法的差异,Sevakula等[21]定义了一个一般化的模糊隶属函数。Liu等[22]提出了一种基于最邻近算法的模糊隶属函数构造方法以降低模糊隶属度的计算复杂度。现有的模糊隶属函数构造方法大都是基于二分类问题。如何设计适应偏好学习中排序问题的模糊隶属函数构造方法仍然是一项挑战。
LiuY, WuF, SunJ, et al.Group recommendation method based on co-evolution of group preference and user preference [J].Syst Eng-Theory Pract, 2021, 41(3): 1-30.
QianM, XuZ.A Study of dynamic recognition of consumer brand decision-making preference based on machine learning method [J].Nankai Bus Rev, 2019, 22(3): 66-76.
DoyleO M, WestmanE, MarquandA F, et al.Predicting progression of Alzheimer’s disease using ordinal regression [J].PLOS One, 2014, 9(8): e105542.
[10]
ChangK, ChenC, HungY.Ordinal hyperplanes ranker with cost sensitivities for age estimation [C]// Proceedings of the 2011 IEEE Conference on Computer Vision and Pattern Recognition.Piscataway: IEEE, 2011: 585-592.
[11]
KimK, AhnH.A corporate credit rating model using multi-class support vector machines with an ordinal pairwise partitioning approach [J].Comput Oper Res, 2012, 39(8): 1800-1811.
[12]
DiazR, MaratheA.Soft labels for ordinal regression [C]// Proceedings of the 2019 IEEE Conference on Computer Vision and Pattern Recognition.Piscataway: IEEE, 2019: 4738-4747.
[13]
PangG, YanC, ShenC, et al.Self-trained deep ordinal regression for End-to-End video anomaly detection [C]// Proceedings of the 2020 IEEE Conference on Computer Vision and Pattern Recognition.Piscataway: IEEE, 2020: 12173-12182.
[14]
UusitaloL.Advantages and challenges of Bayesian networks in environmental modelling [J].Ecol Model, 2007, 203(3/4): 312-318.
[15]
Fdez-DíazL, Fdez-DíazM, QuevedoJ R, et al.Capturing waste collection planning expert knowledge in a fitness function through preference learning [J].Eng Appl Artif Intel, 2021, 99: 104113.
[16]
HerbrichR, GraepelT, Bollmann-SdorraP, et al.Learning preference relations for information retrieval [C]// ICML-98 Workshop: Text Categorization and Machine Learning.Menlo Park: AAAI, 1998: 80-84.
[17]
BrochuE, De FreitasN, GhoshA.Active preference learning with discrete choice data [C/OL]// Advances in Neural Information Processing Systems 21.Laguna Beach: Curran Associates, 2007: 409-416.[2024-03-01].
HeoG, GaderP.Fuzzy SVM for noisy data: A robust membership calculation method [C]// IEEE International Conference on Fuzzy Systems.Piscataway: IEEE, 2009: 431-436.
[25]
SevakulaR K, VermaN K.Compounding general purpose membership functions for fuzzy support vector machine under noisy environment [J].IEEE T Fuzzy Syst, 2017, 25(6): 1446-1459.
[26]
LiuJ, ZioE.A scalable fuzzy support vector machine for fault detection in transportation systems [J].Expert Syst Appl, 2018, 102: 36-43.
[27]
HanssonB.Choice structures and preference relations [M].Berlin: Springer, 1968.
[28]
PiriS, DelenD, LiuT.A synthetic informative minority over-sampling (SIMO) algorithm leveraging support vector machine to enhance learning from imbalanced datasets [J].Decis Support Syst, 2018, 106: 15-29.
[29]
LiuT Y.Learning to rank for information retrieval [M].Berlin: Springer, 2011.
[30]
FanR E, ChangK W, HsiehC J, et al.LIBLINEAR: A library for large linear classification [J].J Mach Learn Res, 2008, 9: 1871-1874.
WangH, ZhengB, YoonS W, et al.A support vector machine-based ensemble algorithm for breast cancer diagnosis [J].Eur J Oper Res, 2018, 267(2): 687-699.
[33]
HerbrichR.Large margin rank boundaries for ordinal regression [C]// Advances in Large Margin Classifiers.Cambridge: MIT Press, 2000: 115-32.
[34]
WuT F, LinC J, WengR C.Probability estimates for multi-class classification by pairwise coupling [J].J Mach Learn Res, 2004, 5: 975-1005.
[35]
LiL, LinH T.Ordinal regression by extended binary classification [C]// Proceeding of the 19th International Conference of Neural Information Processing Systems.Cambridge: MIT press, Curran Associates, 2007: 865-872.
[36]
De RijkM C, LaunerL J, BergerK, et al.Prevalence of Parkinson's disease in Europe: A collaborative study of population-based cohorts.Neurologic diseases in the elderly research group [J].Neurology, 2000, 54(11): S21-S23.
DorseyE R, GeorgeB P, LeffB, et al.The coming crisis: Obtaining care for the growing burden of neurodegenerative conditions [J].Neurology, 2013, 80(21): 1989-1996.
AlmeidaJ S, Rebouças FilhoP P, CarneiroT, et al.Detecting Parkinson’s disease with sustained phonation and speech signals using machine learning techniques [J].Pattern Recogn Lett, 2019, 125: 55-62.
[42]
TsanasA, LittleM A, McSharryP E, et al.Nonlinear speech analysis algorithms mapped to a standard metric achieve clinically useful quantification of average Parkinson's disease symptom severity [J].J R Soc Interface, 2011, 8(59): 842-855.